Creating a personalized avatar (Soul ID) involves going through the following six key steps:
- material preparation stage: Collect at least 10 personal photos, covering a wide range of expressions such as front/side face, smile/seriousness, etc. It is recommended to include the effects of different lighting environments.
- Data upload: Select the Soul ID function in the Diffuse app and upload compliant photos in bulk through the album (the system will automatically check the image quality)
- model training: After submission the cloud GPU cluster starts processing, usually 5-8 minutes to complete feature extraction and dynamic mapping learning (progress bar shows in real time)
- image activation: After training, you can generate a scene with text commands such as "My avatar is having coffee in front of the Paris Tower", which consumes 3 credits the first time you use it.
- Effectiveness Optimization: 20+ parameters can be adjusted in the generation settings, such as style filter (realistic/anime), movement range, light and shadow intensity, etc.
- Application extensions: The finished product can be exported as a 3D model in GLB format or directly inserted into 200+ pre-built template scenes.
Note: It is recommended to use the clear original image taken by the rear camera to avoid beauty filters affecting the modeling accuracy; the copyright of the generated avatar belongs to the creator by default.
This answer comes from the articleHiggsfield AI: Using AI to Generate Lifelike Videos and Personalized AvatarsThe
May not be reproduced without permission:AI productivity tools " How to create a personalized avatar with Higgsfield AI?
